Organoid Models of Human and Mouse Ductal Pancreatic Cancer

SUMMARY

Pancreatic cancer is one of the most lethal malignancies due to its late diagnosis and limited response to treatment. Tractable methods to identify and interrogate pathways involved in pancreatic tumorigenesis are urgently needed. We established organoid models from normal and neoplastic murine and human pancreas tissues. Pancreatic organoids can be rapidly generated from resected tumors and biopsies, survive cryopreservation and exhibit ductal- and disease stage-specific characteristics. Orthotopically transplanted neoplastic organoids recapitulate the full spectrum of tumor development by forming early-grade neoplasms that progress to locally invasive and metastatic carcinomas. Due to their ability to be genetically manipulated, organoids are a platform to probe genetic cooperation. Comprehensive transcriptional and proteomic analyses of murine pancreatic organoids revealed genes and pathways altered during disease progression. The confirmation of many of these protein changes in human tissues demonstrates that organoids are a facile model system to discover characteristics of this deadly malignancy.

INTRODUCTION

Mortality due to pancreatic cancer is projected to surpass that of breast and colorectal cancer by 2030 in the United States (Rahib et al., 2014; Siegel et al., 2013). This dire scenario reflects an aging population, the improvement of outcomes for breast and colorectal cancer patients, the advanced stage at which most patients with pancreatic cancer are diagnosed, and the lack of durable treatment responses in pancreatic cancer patients. Indeed, effective therapeutic strategies for patients with pancreatic ductal adenocarcinoma (PDA) have been difficult to identify (Abbruzzese and Hess, 2014).

The therapeutic resistance of PDA has been explored in a variety of cell culture and animal model systems, with clinically actionable findings encountered only occasionally (Villarroel et al., 2011). Patient derived xenografts (PDXs) have yielded insights into PDA, but their generation requires a large amount of tissue and takes multiple months to establish (Kim et al., 2009; Rubio-Viqueira et al., 2006). Genetically engineered mouse models (GEMMs) of PDA have also been generated as a parallel system to investigate fundamental biology and clinical implications (Perez-Mancera 2012). These GEMMs accurately mimic the pathophysiological features of human PDA, including disease initiation from pre-invasive pancreatic intraepithelial neoplasms (PanINs) (Hingorani et al., 2003; Perez-Mancera et al., 2012), and were used to discover that PDA possesses a deficient vasculature that impairs drug delivery (Erkan et al., 2009; Jacobetz et al., 2012; Koong et al., 2000; Olive et al., 2009; Provenzano et al., 2012). While GEMMs have informed PDA therapeutic development (Beatty et al., 2011; Frese et al., 2012; Neesse et al., 2014), they are expensive and time-consuming (Perez-Mancera et al., 2012). In addition, both human PDA and GEMMs exhibit an extensive stromal component that decreases the neoplastic cellularity, making it difficult to isolate and characterize the epithelium-derived malignant cells in murine pancreatic neoplastic tissues.

To study neoplastic cells, dissociated human tumors are often grown in two-dimensional (2D) culture conditions (Sharma et al., 2010), which do not support growth of untransformed, non-neoplastic pancreatic cells. Three-dimensional (3D) culture strategies have been developed to study normal, untransformed cells, but only allow minimal propagation (Agbunag and Bar-Sagi, 2004; Lee et al., 2013; Means et al., 2005; Rovira et al., 2010; Seaberg et al., 2004). A comprehensive 3D cell culture model of murine and human PDA progression would facilitate investigation of genetic drivers, therapeutic targets, and diagnostics for PDA.

To address this deficiency, we sought to generate normal and neoplastic pancreatic organoids by modifying approaches we previously pioneered to culture intestinal (Sato et al., 2009), gastric (Barker et al., 2010), colon carcinoma (Sato et al., 2011), hepatic (Huch et al., 2013b), pancreatic (Huch et al., 2013a), and prostatic organoids (Gao et al., 2014; Karthaus et al., 2014). We developed 3D organoids from normal and malignant murine pancreatic tissues and used this model system to investigate PDA pathogenesis. Pancreatic organoids derived from wild-type mice and PDA GEMMs accurately recapitulate physiologically relevant aspects of disease progression in vitro. Following orthotopic transplantation, organoids from wild-type mouse normal pancreata are capable of regenerating normal ductal architecture, unlike other 3D model systems. We further developed methods to generate pancreatic organoids from normal and diseased human tissues, as well as from endoscopic needle biopsies. Following transplantation, organoids derived from murine and human PDA generate lesions reminiscent of PanIN and progress to invasive PDA. Finally, we demonstrate the utility of organoids to identify molecular pathways that correlate with disease progression, and that represent therapeutic and diagnostic opportunities.

RESULTS

Murine pancreatic ductal organoids expressing oncogenic Kras recapitulate features of PanINs

Recently, we derived continuously proliferating, normal pancreatic organoids from adult murine ductal cells (Huch et al., 2013a). We optimized this approach to generate models of PDA progression. We manually isolated small intralobular ducts and established organoid cultures from C57Bl/6 mouse normal pancreata and pancreatic tissues that contained low-grade murine PanIN (mPanIN-1a/b) from Kras^+/LSL-G12D; Pdx1-Cre (“KC”) mice (Figure 1A). KC mice develop a spectrum of pre-invasive ductal lesions that mirror human PanINs, and upon aging, stochastically develop primary and metastatic PDA (Hingorani et al., 2003). Ducts from KC pancreata were often larger and exhibited higher grades of dysplasia compared to those from wild-type mice (Figure 1A). After 1–3 days in culture, organoid growth was observed from isolated ducts (Figure 1A). We created a collection of 10 murine normal (mN) and 9 PanIN (mP) organoid cultures that we have continuously propagated for over 20 passages and successfully cryopreserved (Table S1A). mP organoids exhibited recombination of the conditional Kras^LSL-G12D allele and higher levels of Kras-GTP when compared to mN organoids (Figure 1B).

Figure 1. Oncogenic Kras^G12D expression in pancreatic ductal organoids is sufficient to induce pre-invasive neoplasms.

(A) Hematoxylin and eosin (H&E) staining of murine pancreatic tissue used to prepare organoids (top). Arrows indicate mouse normal or PanIN ductal structures. Ducts embedded in Matrigel immediately following isolation (middle) and organoids 3 days post-isolation (bottom). Arrowheads mark isolated ducts and growing organoids. Scale bars, 50 μm.

(B) Immunoblots for Kras, pan-Ras, Kras-GTP by RBD-GST pull-down and Tubulin in mouse normal (mN) and mPanIN (mP) organoids. PCR confirmation of Cre-mediated recombination of the Kras^LSL-G12D allele (bottom).

(C) qRT-PCR of ductal (Pdx1, Ck19, Sox9, Hnf6), acinar (Ptf1a, Cpa1, Amy), and endocrine (Ngn3, Chga, Ins2) lineage markers in mN and mP organoids. Means of 3 biological replicates are shown. Error bars depict standard error of the means (SEMs). Values were normalized to mouse normal pancreas.

(D) qRT-PCR of genes indicative of PanIN lesions (Muc5ac, Muc6, Tff1, Klf4) in mN and mP organoids. Values were normalized to mN organoids. Means of 3 biological replicates are shown. Error bars depict SEMs. **: p < 0.01 by two-tailed Student’s t-test.

(E) H&E, Alcian Blue staining, and immunohistochemistry (IHC) of orthotopic, syngeneic transplants of GFP-transduced mN and mP organoids. Scale bars, 200 μm.

(F) Immunoblots for Kras, pan-Ras, Kras-GTP by RBD-GST pull-down and Tubulin in Kras^+/LSL-G12D organoids transduced with adenoviral-Cre (Ad-Cre) or adenoviral-blank (Ad-Bl). PCR confirmation of Cre-mediated recombination of the Kras^LSL-G12D allele (bottom).

(G) H&E, Alcian Blue staining, and IHC of orthotopic syngeneic transplants of organoids transduced with Ad-Bl (Kras^+/LSL-G12D; R26-LSL-YFP) and Ad-Cre (Kras^+/LSL-G12D; R26-YFP) 2 weeks post-transplant. Scale bars, 200 μm.

See also Figure S1 and Table S1.

To determine the contribution of different pancreatic lineages to the organoids, we evaluated the expression of pancreatic lineage markers in organoids. Genes associated with the ductal lineage (Ck19, Sox9) (Cleveland et al., 2012) were enriched in the mN and mP organoids compared to total pancreatic tissues which contain relatively few ductal cells (Figure 1C). In addition, the mP organoids up-regulated genes indicative of a PanIN disease state (Muc5ac, Muc6, Tff1) relative to mN, with no difference in Klf4 (Figure 1D) (Prasad et al., 2005). GFP-transduced mN and mP organoids were orthotopically transplanted into syngeneic C57Bl/6 or Nu/Nu mice. mN organoids quickly formed ductal structures comprised of simple cuboidal cells that persisted for up to one month (n = 9/27 transplants), but were not observed after 2 months (n = 0/13 transplants) (Figure 1E, Table S1B). In comparison, mP organoids formed small cysts lined with a single layer of simple cuboidal ductal cells interspersed with mucin-containing columnar epithelial cells. Although we could not demonstrate that the mP transplants were contiguous with the native ductal system, they resembled pre-invasive mPanIN (Figure S1C). These dysplastic epithelial cells persisted for 2 months or longer (n = 16/18 transplants), were GFP and Ck19 positive, expressed the mPanIN-associated mucin Muc5ac, and stained prominently with Alcian Blue (Figure 1E, Table S1C). In addition, dysplastic cells had increased proliferation and a robust stromal response when compared to mN transplants, both of which are characteristic of autochthonous mPanIN tissue (Figures S1A–C). The ability of transplanted mP organoids to form lesions with many of the features of mPanINs demonstrates the utility of this system as a model for early pancreatic neoplasia.

Multiple cellular origins have been proposed for the development of PDA, with the pancreatic acinar cell hypothesized to be a major contributor to PDA initiation (De La et al., 2008; Gidekel Friedlander et al., 2009; Guerra et al., 2003; Habbe et al., 2008; Kopp et al., 2012; Morris et al., 2010; Sawey et al., 2007). However, recent studies have suggested that transformation of pancreatic ductal cells can also give rise to PDA (Pylayeva-Gupta et al., 2012; Ray et al., 2011; von Figura et al., 2014). Acinar cells isolated from wild-type pancreata are unable to form organoids in our conditions (Huch et al., 2013a). Therefore, our pancreatic ductal organoid system offers a unique opportunity to determine whether ductal cells can give rise to mPanIN. To assess whether expression of oncogenic Kras in pancreatic ductal organoids is sufficient to induce mPanIN formation in vivo, we derived organoids from ducts harboring the conditional Kras^+/LSL-G12D allele (Hingorani et al., 2003). Following activation of Kras by adenoviral-Cre (Ad-Cre) infection, Kras^G12D organoids maintained expression of genes specific to ductal cells and not acinar or endocrine lineages (Figures S1D and S1E). Recombination of the Kras^+/LSL-G12D allele was confirmed by PCR, and levels of GTP-bound Kras were increased relative to control-infected organoids (Figure 1F). In addition, expression of Kras^G12D resulted in the up-regulation of genes associated with human PanIN (Figure S1F). The Kras^G12D-expressing organoids demonstrated increased proliferation relative to control organoids (Figure S1G). Finally, Kras^G12D organoids formed mPanIN-like structures with columnar cell morphology when implanted orthotopically into syngeneic mice (Figure 1G). This morphology contrasted with the normal-appearing ductal architecture formed by transplanting Kras^+/LSL-G12D organoids or wild-type mN organoids (Figures 1E and 1G). The ability of mPanIN-like structures to develop from Kras^G12D-expressing ductal organoids following transplantation demonstrates that ductal cells are also competent to form mPanINs.

Tumor-derived organoids provide a model for murine PDA progression

We prepared pancreatic ductal organoids from multiple murine primary tumors (mT) and metastases (mM) from KC and Kras^+/LSL-G12D; Trp53^+/LSL-R172H; Pdx-Cre (“KPC”) mice which develop mPDA more rapidly than KC mice (Figure 2A, Table S2A) (Hingorani et al., 2005). mT and mM organoids exhibited recombination of the Kras ^LSL-G12D allele, increased levels of Kras-GTP and Kras protein (Figure 2B). mT and mM organoids had an increased level in S6 phosphorylation, but not Erk or Akt phosphorylation (Figure 2B).

Figure 2. Modeling murine PDA progression with tumor- and metastasis-derived organoids.

(A) H&E staining of murine tissue from which tumor and metastasis organoids were derived (top). Arrowhead indicates metastasis. Scale bars, 50 μm. Digested murine tissues embedded in Matrigel immediately following isolation (middle) and organoids 3 days post-isolation (bottom). Scale bars, 200 μm.

(B) Immunoblots of selected signaling effectors, Kras-GTP and Ras-GTP by RBD-GST pull-down, and Tubulin. PCR confirmation of Kras^LSL-G12D recombination in mP, mT, and mM organoids (bottom).

(C) H&E staining of tumors and metastases (met) derived from mT organoid orthotopic transplants. Scale bars, 200 μm (top) and 50 μm (bottom).

(D) Loss of heterozygosity of the wild-type Trp53 allele determined by PCR (top) and immunoblot analysis of Trp53, Smad4, and p16, and Tubulin. mM3L: derived from a liver metastasis.

(E) Karyotypes of organoids and monolayer (2D) cell lines.

See also Figure S2 and Table S2.

Orthotopic transplantation of mT organoids initially generated low- and high-grade lesions that resembled mPanIN (Figure 2C, Table S2B). Over longer periods of time (1 – 6 months), transplants developed into invasive primary and metastatic mPDA (Figure 2C, Table S2B). mT organoids engrafted with a similar efficiency upon orthotopic transplantation in Nu/Nu mice (91.7%) compared to C57Bl/6 mice (85%), but disease progression was accelerated in Nu/Nu hosts (Table S2B). While most mT organoid transplants required several months to progress from early mPanIN-like lesions to invasive and metastatic cancer (Figure 2C, Table S2B), mM organoids rapidly formed invasive mPDA within 1 month (Table S2C). The ability of organoid transplants to reproduce the discrete stages of disease progression contrasts with the rapid formation of advanced mPDA following transplantation of 2D cell lines (Figure S2A–C) (Olive et al., 2009).

The tumor exhibited a prominent stromal response in the transplanted mT and mM organoids and resembled autochthonous tumors from KPC mice (Figure S2A) (Olive et al., 2009). This stromal response is often absent in tumors formed from 2D cell lines (Figure S2A) (Olive et al., 2009). Low vascular density and high vessel-to-tumor distance were also observed, demonstrating the close resemblance of the organoid transplantation models to autochthonous mPDA, in contrast to transplanted 2D cell lines (Figure S2A–C) (Olive et al., 2009).

Loss of heterozygosity (LOH) for Trp53 has been reported as a common feature of mPDA based on studies of 2D cell lines (Hingorani et al., 2005). Therefore, we assayed for Trp53 LOH in our murine 3D organoids. All mT organoids prepared from KPC tumors maintained expression of p16, did not exhibit Trp53 LOH and maintained a stable karyotype, whereas most mM organoids lost the wild-type Trp53 allele and were aneuploid (Figures 2D, 2E and S2D). We generated 2D cell lines from mT and mM organoids, but found that mN and mP organoids were unable to propagate in 2D. mT1 was derived from a KC mouse PDA, lacks the mutant Trp53 allele and was also unable to propagate in 2D. All mT-derived 2D cell lines exhibitedTrp53 LOH and were aneuploid (Figure 2E and S2D).

To determine whether organoids are suitable for genetic cooperation experiments, shRNAs targeting p53 and p16/p19 were introduced into mP organoids (Figure S2E). While the proliferation of mP organoids increased upon knockdown of either p53 or p16/p19 (Figure S2G), only p53 knockdown enabled 2D growth and colony formation (Figure S2F, data not shown). Also, only p53 knockdown promoted progression of mP organoid transplants to invasive carcinoma within 3 months (Figure S2H). This contrasts with a previous report that biallelic loss of p16/p19 and Kras mutation promoted mPDA (Aguirre et al., 2003; Bardeesy et al., 2006), and may reflect differences in the genetic system or the initiating cellular compartment. Nevertheless, the cooperation between p53 depletion and oncogenic Kras demonstrates that organoids are a facile system to evaluate genetic mediators of PDA progression.

Human pancreatic organoids model PanIN to PDA progression

We modified our culture conditions to support the propagation of human normal and malignant pancreatic tissues. Isolation of ductal fragments was not always feasible because some normal pancreatic tissue samples were pre-digested in preparation for islet transplantation. Therefore, we directly embedded digested material into Matrigel. This approach achieved an isolation efficiency of 75–80% for human normal (hN) organoids (Figure 3A and S3, Table S3). hN organoids require TGF-beta pathway inhibitors (A83-01 and Noggin), R-Spondin1 and Wnt3a-conditioned media, EGF, and PGE2 for propagation (Figures 3B and 3C). Unlike mN organoids, which have unlimited propagation in culture, hN organoids ceased proliferating after 20 passages or approximately 6 months, but could be cryopreserved.

Figure 3. Human pancreatic ductal organoids recapitulate features of normal and neoplastic ducts.

(A) Representative images (top) and H&E staining (middle) of human organoid cultures established from: normal tissues (hN1-2), resected primary tumors (hT1-2), a resected metastatic lung lesion (hM1) and a fine-needle aspiration biopsy of a metastatic lesion (hFNA2). H&E staining of the resected tissues from which the organoids were derived (bottom). Scale bars, 500 μm (top), 250 μm (middle), and 500 μm (bottom).

(B) Representative images of hN and hT organoids cultured for two weeks (1 passage) in human complete media, or in human complete media lacking the indicated factors. Scale bars, 500 μm.

(C) Number of passages hN and hT organoids could be propagated in the absence of the indicated factors.

(D) Targeted sequencing analysis of human organoids. Genes altered in more than one sample and/or known to be mutated in PDA are shown. If multiple mutations were found in a gene, only one mutation per gene is shown. Color key for the type of genetic alterations is shown. Met indicates organoids derived from metastatic samples.

See also Figure S3, Table S3 and S4

We modified the methods described above to accommodate the extensive desmoplastic reaction in freshly resected PDA specimens and generated human tumor-derived organoids (hT) (Figures 3A and S3). hT organoids could be passaged indefinitely and cryopreserved (Figure 3C). The establishment hT organoids had efficiencies of 75% (n = 3/4) and 83% (n = 5/6) in the Netherlands and USA, respectively (Table S3). The first specimen that failed to generate an organoid culture was obtained from a patient that had undergone neo-adjuvant chemotherapy, and histologic examination of this specimen revealed extensive necrosis. The second specimen that did not generate an organoid culture was predominantly composed of stromal cells, without sufficient viable tumor cells to establish a culture. While the hN organoids had a simple, cuboidal morphology, the hT organoids had differing degrees of dysplastic tall columnar cells, resembling low-grade PanINs (Figures 3A). hT organoids tolerated the withdrawal of certain growth factors from the media (Figures 3B and 3C).

85% of pancreatic cancer patients are ineligible for surgical resection of their tumors (Ryan et al., 2014). Therefore, we determined whether hT organoids could be generated from the limited amount of cellular material provided by endoscopic biopsies using fine needle aspirations (FNA). Initial attempts to generate organoids from FNA biopsies were hampered by loss of cellular material during digestion. Upon optimization of these conditions, human FNA biopsy organoids (hFNA) were generated from two specimens that were not dissociated prior to suspension in Matrigel (Figures 3A and S3). This approach is broadly applicable to PDA patients and enables serial sampling.

Targeted sequencing of 2,000 cancer-associated genes was performed on hN and hT organoids. As expected, no mutations were detected in the hN organoid cultures. These analyses identified oncogenic KRAS mutations in the majority of tumor-derived samples (n = 8), and mutations in TP53 (n = 7), SMAD4 (n = 5), and CDKN2A (n = 4) (Figure 3D, Table S4). We also noted amplification of known oncogenes, such as MYC (n = 4), and loss of tumor suppressors, including TGFBR2 (n = 3) and DCC (n = 5). Importantly, the same KRAS mutations observed in several hT organoids were confirmed in the primary PDA from which they were derived (Table S4). The allele frequency of oncogenic KRAS variants in hT1 – hT5 and hFNA2 ranged from approximately 50 - 100%. In contrast, the KRAS^G12V allele frequency in hFNA1 was only 1% (Table S4), which may result from co-existence of wild-type ductal cells. While KRAS mutations were not detected in hT8 (Figure 3D, Table S4), the presence of mutations in known PDA genes (ARID1A and MLL3) suggests that hT8 contains malignant cells despite the absence of a KRAS mutation (Table S4).

To further characterize the cell types present in primary PDA organoids, we evaluated the expression of pancreatic lineage markers. hN and hT organoids expressed markers of ductal cells but not other pancreatic lineages (Figure 4A). The karyotypes of hT organoids were highly aneuploid, whereas the hN organoids were predominantly and stably diploid (Figure 4B). The PDA-associated biomarker CA19-9 (Makovitzky, 1986) was also elevated in hT relative to hN organoids (Figure 4C). The hN and hT organoids are therefore reflective of normal and neoplastic human pancreatic ductal cells and offer a model system to explore pancreatic cancer biology in the more genetically complex background of human cancer.

Figure 4. Molecular characterization and orthotopic transplantation of human organoids.

(A) qRT-PCR of pancreas lineage markers in hN (n = 3) and hT (n = 4) organoids. Mean expression levels were normalized to total pancreas. Error bars depict SEMs.

(B) Karyotyping of human organoids (2 hN, 2hT) at the indicated passages (P).

(C) CA19-9 and actin levels in hN, hT, or hM organoids. The solid line indicates non-congruent lanes.

(D) H&E, Alcian Blue staining and IHC of orthotopic hT2 transplants and the primary tumor. Scale bars, 200 μm (top two panels) and 50 μm (lower two panels).

See also Figure S4 and Table S4D

Following orthotopic transplantation into Nu/Nu mice, hN organoids produced normal ductal structures at low efficiency (n = 2/23), while hT organoids efficiently generated a spectrum of low and high-grade, extra-ductal PanIN-like lesions within one month (n = 9/12) (Figures 4D, S4A, Table S4). The hT-derived transplants initially formed well-defined hollow lesions lined by a single layer of columnar epithelial cells with apical mucin, and basally located, relatively uniform nuclei. The nuclei were small and lacked the pleomorphism and hyperchromasia often seen in invasive PDA. These lesions progressed over several months to infiltrative carcinoma comprised of poorly defined and invasive glands (Figures 4D, S4A, Table S4). A prominent desmoplastic reaction was present in hT-derived PanIN-like structures and PDA, including the deposition of a collagen-rich stroma and the recruitment of αSMA-positive cells (Figure S4B). The mutation or loss of TP53 or SMAD4 in hT1 and hT2 was also detected by IHC in these tumors (Figure S4C, Table S4). Overall, hT organoids represent a transplantable model of human pancreatic cancer progression.

Gene expression analysis of murine pancreatic ductal organoids implicates candidate genes in PDA progression

The mouse organoids were prepared from syngeneic mice, offering the ability to discern gene expression changes in organoids and determine whether these changes correlate with PDA progression. We harvested RNA from mN (n = 7), mP (n = 6), and mT (n = 6) organoids and generated strand-specific RNA-sequencing (RNA-seq) libraries. Sequences were mapped to the mm9 version of the mouse genome, and relative transcript abundances (transcripts per million) of 29,777 mouse genes were determined (Table S5). Principal component analysis revealed that mN organoids were distinct from mP and mT organoids (Figure 5A and Table S5).

Figure 5. Gene expression analysis of murine organoids reveals genetic changes correlated with pancreatic cancer progression.

(A) Principal component analysis of gene expression data for mN, mP, and mT organoids.

(B) The number of genes differentially expressed (DESeq adjusted p value < 0.05) among mN (n = 7), mP (n = 6), and mT (n = 6) organoids.

(C) Heatmap showing relative expression levels using Z-score normalization among mN, mP, and mT organoids. Color key of Z-score is shown.

(D) Venn diagrams show overlap of genes significantly differentially expressed in mP and mT relative to mN organoids. The p values for overlaps were determined by two-tailed Fisher’s Exact test.

(E) Genes with the largest fold-changes in mP or mT relative to mN organoids.

(F) qRT-PCR validation of mN, mP and mT organoid gene expression changes. Values were normalized to mean levels in mN organoids. n = 8 mN, 7 mP, and 8 mT organoid cultures. Error bars show SEMs. *, **, ***, ns: p < 0.05, 0.01, 0.001, or not significant by two-tailed Student’s t-test.

See also Figure S5 and Table S5.

Genes whose levels differed significantly among mN, mP, and mT organoids were identified. 772 genes were found down-regulated and 863 genes up-regulated in mP relative to mN organoids (Figure 5B and Table S5). When mT organoids were compared to mN organoids, 2,721 genes were down-regulated and 2,695 up-regulated. In addition, 823 genes were down-regulated and 640 genes up-regulated in mT relative to mP organoids. Distinct patterns of gene expression were found in the dataset (Figure 5C). The majority of genes differentially expressed in mP relative to mN organoids changed in a similar manner in mT relative to mN organoids (Figure 5D). However, a much larger cohort of genes changed in expression in mT relative to mN than in mP relative to mN organoids (Figure 5D), suggesting that mP organoids represent an intermediate state between mN and mT organoids.

The glycosyltransferase Gcnt3 and putative protein disulfide isomerase Agr2 were among the most up-regulated genes in both mP and mT organoids, and have been demonstrated to be elevated in human PDA (Figure 5E) (Dumartin et al., 2011; Zhao et al., 2014). The most up-regulated gene in both mP and mT relative to mN organoids was the acyl-CoA synthetase Acsm3 (Figure 5E). RNA-seq results were confirmed by qRT-PCR for 35 out of 40 genes (Table S5), including the up-regulation of Agr2, Acsm3, Gcnt1, Gcnt3, and Ugdh, and the down-regulation of the Ptprd gene in mP and mT organoids (Figure 5F and Table S5). Among the genes up-regulated in mP and mT relative to mN organoids, Gcnt1, Gcnt3, Acsm3, Agr2, Syt16, Nt5e, and Ugdh were up-regulated following the Ad-Cre induced expression of oncogenic Kras^G12D, suggesting that these genes are activated downstream of mutant Kras^G12D (Figure S5A). To determine whether organoid RNA-seq profiles resembled gene expression patterns in vivo, we compared our organoid RNA-seq data to a published transcription profile of murine pancreatic tumors upon Kras^G12D inactivation (Ying et al., 2012). Genes differentially expressed upon oncogenic Kras activation overlapped significantly with those up or down-regulated in mP or mT relative to mN organoids (Figure S5B). These analyses demonstrate the ability of the organoid system to identify molecular alterations associated with PDA progression.

Proteomic alterations in murine pancreatic ductal organoids predict pathways associated with PDA progression

As an orthogonal method to investigate molecular alterations in murine pancreatic organoids, we characterized the global proteome of mN (n = 5), mP (n = 4) and mT (n = 5) organoids. Protein lysates were processed using amine-reactive isobaric tags for relative and absolute quantification (iTRAQ) mass spectrometry (Wiese et al., 2007). Samples were run in four 8-plex experiments and merged using an approach that normalizes the data to common samples included across all experiments (Supplemental Experimental Procedures). Upon merging, 6,051 unique protein isoforms were quantified in all samples. We applied linear regression modeling on the normalized intensity peak values and identified 710 protein isoform expression changes between mN and mP organoids (Figure 6A). 1,047 protein isoforms changed expression between mN and mT organoids, and 63 differentially expressed proteins were identified between mP and mT (Figure 6A). The relatively small number of protein expression changes identified between mP and mT organoids reflects the biological similarity between mP and mT (Figure S6A).

Figure 6. Proteomic profiling of murine organoids uncovers molecular pathways linked to pancreatic cancer progression.

(A) Protein expression changes by iTRAQ proteomic analysis of murine organoids. Both unique protein isoforms and protein isoforms encoded by the same gene are included (adjusted p value < 0.1 by linear regression analysis).

(B) Heatmap of unique protein isoforms that differ (adjusted p value < 0.05) among mN, mP, and mT organoids. Color key of the Z-score is shown.

(C) Venn diagrams showing overlaps between proteins differentially expressed (p < 0.05) in mP and mT relative to mN organoids. p values for overlaps were determined by two-tailed Fisher’s Exact test.

(D) Venn diagrams showing overlaps between genes and proteins found differentially expressed by RNA-seq and proteomic analyses (adjusted p < 0.05). p values for the overlaps were determined by two-tailed Fisher’s Exact test.

(E) Molecular pathways found enriched by GSEA analysis of RNA-seq and proteomic data. Normalized enrichment scores (NESs), p and q values are shown.

(F) Heatmap showing relative gene expression levels of nucleoporins in mN, mP, and mT organoids determined by RNA-seq. Color key of the Z-score is shown.

See also Figure S6, Table S6 and S7.

mN organoids showed unique proteomic profiles from their mP and mT counterparts (Figure 6B and 6C). To compare the proteomic and RNA-seq data, we collapsed the unique protein isoforms into their corresponding 4,155 genes. Some protein expression changes (e.g. 123/150 for down-regulated and 96/151 for up-regulated mP proteins) did not reflect corresponding transcriptional changes, indicating that protein stability may play a role in cancer progression, particularly in mP organoids (Figure 6D). Nonetheless, the proteomic data validated many of the expression changes identified by RNA-seq (Figure 6D), including up-regulation of Gcnt3, Agr2, and Ugdh (Table S6). Additionally, of the 1,599 genes whose expression levels changed in mT relative to mN organoids that were measured by mass spectrometry, 301 (19%) showed corresponding protein changes (Figure 6D).

Gene Set Enrichment Analysis (GSEA) on the RNA-seq and proteomic data (Subramanian et al., 2005) revealed elevated expression of genes and proteins involved in Glutathione Metabolism and Biological Oxidations in mP relative to mN organoids (Figures 6E, S6B, S6C, Table S7), consistent with elevations in reactive oxygen species metabolism previously reported in Kras^G12D cells (DeNicola et al., 2011; Ying et al., 2012). The enrichment of proteins involved in Glutathione metabolism was also found in mT relative to mN organoids (Table S7). Additionally, we identified a significant positive enrichment of proteins involved in the Steroid Biosynthesis, Cholesterol Biosynthesis, One Carbon Pool by Folate, and Pyrimidine Metabolism pathways (Figures 6E, S6B, S6C, Table S7), consistent with an earlier report (Ying et al., 2012). Similar pathways were enriched in mP relative to mN organoids (Cholesterol Biosynthesis, One Carbon Pool by Folate, and Pyrimidine Metabolism) (Figures S6B and S6C, Table S7), while Fatty Acid Metabolism and TCA Cycle/Respiratory Electron Transport pathways were down-regulated (Figures S6B and S6C, Table S7). The increase in anabolic and decrease in catabolic pathways suggest that complex alterations in fatty acid metabolism occur during PDA progression.

Interestingly, we also found broad up-regulation of the nucleoporin family at both the RNA and protein levels in the mT relative to mN organoids (Figure 6F, Table S6). The individual nucleoporins NUP214, NUP153 and NUPL1 were previously identified in shRNA dropout screens in pancreatic cancer cell lines (Cheung et al., 2011; Shain et al., 2013). Furthermore, amplification of NUP153 was detected in one human PDA cancer cell line, and elevation of NUP88 was detected in human primary PDA (Cheung et al., 2011; Gould et al., 2000; Shain et al., 2013). This systematic analysis of molecular alterations in pancreatic organoids implicates nuclear transport as a pathway correlated with pancreatic cancer progression.

In vivo mouse and human validation of candidates associated with PDA progression in organoids

To demonstrate that the mouse organoid culture system represents a biological resource for the accurate discovery of genes associated with PDA progression, we selected 16 genes up-regulated in mT organoids for validation in primary tissue specimens by IHC and immunofluorescence (IF) (Figure 7A). These 16 genes included enzymes, membrane proteins, structural proteins and secreted ligands, which could represent candidate biomarkers and therapeutic targets. Of the 14 antibodies that generated a detectable signal on murine pancreatic tissue sections, 13 antibodies confirmed the increased expression of the candidate protein in mPanIN and mPDA lesions in concordance with the RNA-seq and proteomic data (Figures 7A, 7B and S7A). 11 of the 13 candidate antibodies were compatible for evaluation in human tissues, and 7 of these candidates were up-regulated in human PDA when compared to normal pancreatic ductal tissues (Figures 7C, 7D and S7B). The high expression of many of these markers was recapitulated in orthotopic transplants of hT organoids into Nu/Nu mice (Figure 7C). These results indicate that the organoid culture system accurately models PDA progression and can serve as a resource for the discovery and genetic dissection of pathways driving human pancreatic tumorigenesis.

Figure 7. Increased levels of ACSM3, NT5E and GCNT3 correlate with mouse and human PDA progression.

(A) IHC analysis of 14 candidate genes in mouse adjacent normal ducts, mPanlN and mPDA. Differential expression is indicated as: - (negative), + (weak), ++ (moderate) or +++ (strong).

(B) IHC analysis of ACSM3, NT5E and GCNT3 in mouse normal ducts, mPanlN and mPDA tissues. Arrow indicates adjacent normal ducts in mPanlN tissues. Arrowhead indicates mPanlN or mPDA. Scale bars, 50 μm.

(C) IHC analysis of 7 candidate genes in human normal pancreas, hT orthotopic transplants and PDA tissues. Differential expression is indicated as: - (negative), + (weak), ++ (moderate) or +++ (strong). Only the ductal component of the normal pancreas was scored.

(D) IHC analysis of ACSM3, NT5E and GCNT3 in human normal pancreas and PDA tissues. Arrow indicates normal ducts and arrowhead indicates PDA. Scale bars, 50 μm.

See also Figure S7.

DISCUSSION

We have established pancreatic organoids as a tractable and transplantable system to probe the molecular and cellular properties of neoplastic progression in mice and humans. In contrast to prior reports (Agbunag and Bar-Sagi, 2004; Rovira et al., 2010; Seaberg et al., 2004), our culture conditions prevent the rapid exhaustion of normal ductal cells in vitro, and generate a normal ductal architecture following orthotopic transplantation. Importantly, the ability to passage and transplant both normal and neoplastic ductal cells enables a detailed analysis of molecular pathways and cellular biology that is not possible when neonatal pancreatic fragments are propagated in air-liquid interfaces, or induced pluripotent cells are employed (Agbunag and Bar-Sagi, 2004; Kim et al., 2013; Li et al., 2014). For example, our finding that nucleoporins are broadly up-regulated in the neoplastic murine organoids, coupled with the known associations of nucleoporins to cell proliferation and cell transformation, presents a class of proteins to investigate in pancreatic cancer progression (Gould et al., 2000; Kohler and Hurt, 2010). Furthermore, the ability to systematically characterize human pancreatic cancer organoids that lack KRAS mutations, such as hT8, will reveal driver genes for PDA. Finally, since organoids can be readily established from small patient biopsies, they should hasten the development of personalized approaches for pancreatic cancer patients.

EXPERIMENTAL PROCEDURES

Animals

Trp53^+/LSL-R172H, Kras^+/LSL-G12D and Pdx1-Cre strains in C57Bl/6 background were interbred to obtain Pdx1-Cre; Kras^+/LSL-G12D (KC) and Pdx1-Cre; Kras^+/LSL-G12D; Trp53^+/LSL-R172H (KPC) mice (Hingorani et al., 2005). The R26^LSL-YFP strain was interbred to get the desired genotype. C57Bl/6 and athymic Nu/Nu mice were purchased from Charles River Laboratory and Jackson Laboratory. All animal experiments were conducted in accordance with procedures approved by the IACUC at Cold Spring Harbor Laboratory (CSHL).

Murine pancreatic ductal organoid culture and analysis

Detailed procedures to isolate normal pancreatic ducts have been described previously (Huch et al., 2013a). In brief, normal and pre-neoplastic pancreatic ducts were manually picked after enzymatic digestion of pancreas with 0.012% (w/v) collagenase XI (Sigma) and 0.012% (w/v) dispase (Gibco) in DMEM media containing 1% FBS (Gibco), and were seeded in growth factor-reduced (GFR) Matrigel (BD). For tumors and metastases, bulk tissues were minced and digested overnight with collagenase XI and dispase and embedded in GFR Matrigel. All RNA-seq data are available at GEO under accession number GSE63348. The proteomic raw data are available at PeptideAtlas under accession number PASS00625.

Human specimens

Pancreatic cancer tissues and adjacent normal pancreas were obtained from patients undergoing surgical resection at the University Medical Centre Utrecht Hospital, Memorial Sloan-Kettering Cancer Center (MSKCC), MD Anderson Cancer Center (MDACC), and Weill Cornell Medical College (WCMC). Normal pancreatic tissue was also obtained from islet transplant programs at the University of Illinois at Chicago and University of Miami Miller School of Medicine. All human experiments were approved by the ethical committees of the University Medical Centre Utrecht or the IRBs of MSKCC, MDACC, WCMC and CSHL. Written informed consent from the donors for research use of tissue in this study was obtained prior to acquisition of the specimen. Samples were confirmed to be tumor or normal based on pathological assessment.

Human pancreatic tumor and normal organoid culture and analysis

Tumor tissue was minced and digested with collagenase II (5 mg/mL, Gibco) in human complete medium (see below) at 37°C for a maximum of 16 hours. The material was further digested with TrypLE (Gibco) for 15 minutes at 37°C, embedded in GFR Matrigel, and cultured in human complete medium [AdDMEM/F12 medium supplemented with HEPES (1x, Invitrogen), Glutamax (1x, Invitrogen), penicillin/streptomycin (1x, Invitrogen), B27 (1x, Invitrogen), Primocin (1mg/ml, InvivoGen), N-acetyl-L-cysteine (1 mM, Sigma), Wnt3a-conditioned medium (50% v/v), RSPO1-conditioned medium (10% v/v, Calvin Kuo), Noggin conditioned medium (10% v/v) or recombinant protein (0.1 μg/mL, Peprotech), epidermal growth factor (EGF, 50 ng/ml, Peprotech), Gastrin (10 nM, Sigma), fibroblast growth factor 10 (FGF10, 100 ng/ml, Prepotech), Nicotinamide (10 mM, Sigma) and A83-01 (0.5 μM, Tocris)].

Normal samples were processed as above, except that the collagenase digestion was done for a maximum of 2 hours in the presence of soybean trypsin inhibitor (1 mg/ml, Sigma). Following digestion, cells were embedded in GFR Matrigel and cultured in human complete medium with the addition of PGE2 (1 μM, Tocris). The targeted DNA-sequencing data are available at Ensembl under the accession number ERP006373.

Additional experimental details and methods can be found in the Supplemental Extended Experimental Procedures.